hi all
We have a study in which participants view three kinds of video in the scanner. Because of the nature of the stimuli it was not possible to equate them in length, so instead they range from about 25-40 seconds. I have modelled them using -stim_times_AM1 with dmUBLOCK following Bob's recommendation (as I understand it) in the 3dDeconvolve help.
While the results are fine, I am curious about the shape of the responses, especially at the start of the videos. I tried using 'TENTzero(-3,25,11)' and got reasonable-looking results but I am not sure how reliable they are since I am leaving a lot of unmodelled stimulus. Does anybody have any advice about this?
Secondly, in the GLM using dmUBLOCK the responses in the left ATL come out as C > A=B, where the difference between C and B is significantly different from that between A and B. However, when I look at the responses that come out of the TENT GLM, the response to A is higher than that to B at its peak, though the trough-to-peak distance is greater for B (see attached picture if it attached, though it does not look like it has). In other words, if the blue line (A) were moved down to line up with the other two at the trough around point 3 on the x-axis, then the responses would fit what comes out of the GLM using dmUBLOCK. But I can't just subtract from one series because I feel like it. So how do I square what I get from dmUBLOCK with what I get from TENT? Or am I trying to square a circle, or comparing apples and oranges, or whatever?
Sorry for the long post. Thanks in advance.
James